graph neural networks

Bacteria tracking and life cycle state classification using graph neural networks and pretrained vision transformers.

For understanding dynamic biological processes such as the life cycle progression of bacteria at the single-cell level, automatic methods for tracking and state classification are needed. In this work, we propose a unified framework based on Graph Neural Networks for simultaneous bacteria tracking, division detection, and life cycle state classification. In previous work, these tasks were treated separately. With our method, trajectories are represented by a graph, where nodes represent bacteria at different time points of a live-cell microscopy video, and edges represent their interactions over multiple frames. Tracking, division detection, and life cycle state classification are performed simultaneously by classifying graph nodes and edges. For all three tasks, we use visual object features from a large-scale pretrained foundation model. This eliminates the need for separately-trained task-specific CNN encoders as used in previous work and enhances the robustness. In addition, we introduce a network-based approach for segmentation error correction using division and multi-frame correspondence predictions. Our method was evaluated using live-cell bright-field microscopy videos of spore germination and outgrowth of rod-shaped bacteria. Our experiments show that the proposed method outperforms existing methods for division detection and state classification. The method yields state-of-the-art results for bacteria tracking and shows increased robustness against segmentation errors as well as image distortions.

PROTEUS: A 40 nm Programmable General-Purpose Digital Compute-In-Memory Accelerator With eNVM and Hierarchical ISA for Versatile Edge AI

We present PROTEUS, an 18 mm2 programmable general-purpose digital compute-in-memory (GP-DCIM) accelerator integrating 4 Mb resistive random access memory (RRAM) and 2.6 Mb tensor static random access memory (SRAM) with a 32-bit hierarchical DCIM instruction set architecture (ISA). PROTEUS features fine-grained 1-D matrix tiling and a reconfigurable DCIM datapath/pipeline for near-100% memory utilization, supporting INT8/INT16/FP8/FP16 DCIM computations. PROTEUS unifies SRAM/RRAM dataflows and embeds nonvolatile micro-programs in RRAM to enable rapid switching among prestored kernels without incurring off-chip instruction feeds or RRAM rewrites. Fabricated in 40 nm ultra-low power (ULP) CMOS with foundry RRAM, PROTEUS delivers 702 GOPS throughput, 6.4 TOPS/W energy efficiency, and 0.039 TOPS/mm2 compute density. It is validated on ResNet-20, BERT-Tiny, MobileViT, GraphSAGE, and Vision Mamba, demonstrating versatility across CNN, Transformer, hybrid CNN-Transformer, graph neural network (GNN), and state-space model (SSM) workloads.

MADGCN: A Meteorology-Aware Spatio-Temporal Graph Convolution Network for Long-Term Air Pollution Forecasting

Air quality forecasting has attracted increasing attention as global air pollution worsens. Spatiotemporal graph neural networks have become a leading paradigm, thanks to their ability to capture complex spatial and temporal dynamics in Air Quality Index (AQI) data. However, existing methods remain limited by weak modeling of long-range temporal dependencies and insufficient integration of meteorological factors. Building on a publicly available nationwide air quality dataset spanning eight years, we propose MADGCN, a Meteorology-Aware Decoupled Spatio-Temporal Convolutional Network that jointly addresses long-horizon temporal modeling and meteorological context fusion. MADGCN includes a dynamic causality discovery module grounded in Granger causality, which captures time-varying causal relationships between meteorological conditions and AQI dynamics. The inferred causal structures further guide a causal graph convolution module and a PatchMixer module, enabling effective spatial interaction modeling and multiscale temporal dependency learning. Extensive experiments against 16 strong baselines show that MADGCN achieves competitive performance for long-horizon air pollution forecasting and generalizes well under high-pollution regimes..

MOSAIC: Multigranularity OOD Detection for IoT Networks via Self-Aligned In-Distribution Consistency

Out-of-distribution (OOD) anomalies pose serious threats to the reliability and security of Internet of Things (IoT) systems. As graph neural networks (GNNs) have become a dominant framework for modeling the relational structures inherent in IoT networks, detecting OOD nodes on graphs has emerged as an essential requirement for trustworthy IoT deployment. For graph node-level OOD detection, current leading methods predominantly follow the OOD-exposure paradigm, which leverages real or synthesized OOD samples to explicitly separate in-distribution (ID) and OOD scores during training. However, this paradigm suffers from two fundamental limitations: 1) real OOD samples are inherently difficult to collect, as novel anomaly patterns in evolving IoT environments are unpredictable before deployment and 2) synthesized pseudo-OOD samples inevitably deviate from true OOD samples, introducing a distributional mismatch that undermines detection reliability. Moreover, both routes rely on regularization hyperparameters that cannot be validated without OOD data. Together, these limitations render OOD-exposure methods illsuited for real-world IoT deployments. To address these issues, we propose multigranularity OOD detection via self-aligned ID consistency (MOSAIC), a graph node-level OOD detection framework that avoids reliance on OOD samples and instead characterizes the ID distribution itself from multiple complementary granularities. MOSAIC evaluates each node from three perspectives: macrolevel global deviation via cosine distance to a distance-weighted ID centroid, mesolevel class-aware boundary deviation via minimum Euclidean distance to class-specific centroids, and microlevel representational stability via feature-masked embedding consistency. A homoscedastic uncertainty framework further balances the training objectives automatically, eliminating the manually tuned regularization hyperparameters required by OOD-exposure methods. Extensive experiments on five social-IoT proxy benchmarks and a bitcoin transaction graph demonstrate that MOSAIC matches or surpasses OOD-exposure-based methods while requiring no OOD data during training, offering a practical solution for open and evolving graph-based IoT systems. The codes are available at https://github.com/Brucesustech/MOSAIC

Integrated Fault Location Method Using Feeder-Centric Graph Neural Networks for Renewable-Penetrated DC Distribution Networks

This paper proposes a fault location method for DC distribution networks (DCDNs) based on graph neural networks (GNNs), which integrates the fault line selection (FLS) and fault distance estimation (FDE) that are conventionally handled independently. The proposed method focuses on the analysis of feeders, including FLS of multiple feeders and FDE of a single feeder. Specifically, a feeder-as-node graph is constructed, where synchronous measurement data are extracted as node features, ensuring consistent dimensionality and enhanced learning efficiency. Moreover, the proposed method explicitly embeds the logical interdependencies between FLS and FDE into the structural design and parameter updating mechanism. An output processing module is designed to estimate the fault distance by analyzing the FLS results, ensuring the model utilizes the data of the DCDN system-level information rather than a single feeder. Furthermore, a two-stage pre-training strategy is introduced to improve stability and generalization, in which partial parameters are frozen. The effectiveness and generalization of the proposed method are verified by hardware-in-the-loop experiments.

GNN-Transformer for Real-Time Power-Constrained Active RIS Configuration in Terahertz Communications

Terahertz (THz) communication is an essential component of sixth-generation (6G) wireless networks; however, its use is constrained by molecular absorption, ultrawideband beam squint, high- $\kappa $ Rician fading, and beam misalignment. The multiplicative fading penalty of passive RIS design can be avoided, and reliable coverage can be extended into the THz band using active, reconfigurable intelligent surfaces (A-RISs) that integrate per-element amplifiers. However, the simultaneous optimization of discrete phase shifts and continuous per-element amplification factors is a nonconvex, high-dimensional problem that is difficult to solve with conventional iterative solvers for the submillisecond coherence times of mobile THz users. A hardware-aware A-RIS-assisted THz communication framework is presented in this article, with an extensive channel model that accounts for frequency-selective molecular absorption, ultrawideband beam squint, high- $\kappa $ Rician fading ( $\kappa \in [{10,25}]$ dB), and stochastic beam misalignment. We present an unsupervised graph neural network-Transformer (GNN-Transformer) architecture to solve the resulting joint optimization problem in real time. Graph convolutional layers exploit local spatial dependencies among RIS elements, while Transformer attention layers capture global channel dependencies across the entire RIS panel. The network is trained end-to-end on channel realizations but, unlike optimal labeling, does not require optimal discrete-phase and continuous-amplification configurations, which are obtained in a single forward pass. The results of the simulations show that the proposed method, which performs an exhaustive search, achieves a higher SNR, meets the power budget without violating it, and reduces inference latency compared to successive convex approximation (SCA), thereby enabling large-scale deployments of THz A-RIS in real time.

Interference Sensing-as-a-Service: End–Edge–Cloud IoT System With Zero-Shot Detection and GNN Localization

Urban Internet of Thing (IoT) networks face severe reliability threats from diverse wireless interference, including jamming and spoofing, which are difficult to detect and localize in multipath-rich environments. Existing schemes often suffer from high false alarms, poor generalization, and low localization accuracy. This article presents an end–edge–cloud interference detection and localization framework integrating zero-shot detection, game-theoretic collaborative sensing, and GNN-based localization. Experiments on a city-scale prototype show that the system achieves 98.5% detection accuracy with recall of 96.7%, while reducing false alarms to 3.2%. The proposed graph neural network (GNN) reduces median localization error to 12.3 m, significantly outperforming baseline methods. Furthermore, the architecture reduces energy consumption by nearly 40% compared with cloud-only designs and maintains end-to-end latency under 120 ms. These results demonstrate that the proposed system enables robust, real-time interference awareness for large-scale IoT deployments, paving the way toward resilient 6G smart cities.

Physics-Informed Series-Aware Graph Transformer Model for Net Load Forecasting

The growing integration of renewable energy sources (RESs), such as photovoltaic (PV) and wind (WD), has significantly increased the variability of net load (NL), posing critical challenges on the net load forecasting. In this paper, a novel physics-informed series-aware graph Transformer (PISAGT) model is proposed for the net load forecasting, which synergistically combines the advantages of both direct and indirect forecasting methods to achieve superior forecasting performance. Firstly, a physics-informed loss module (PILM) is proposed, which introduces the physical characteristics of the net load into the loss function to incorporate physical knowledge in the gradient descent optimization, thereby enhancing the model’s generalization capability for more stable and trustworthy predictions. Secondly, a variable-scale patch embedding method segments time series into subsequence-level patches and transforms them into 2D representations, which enables the forecast model to simultaneously capture local temporal patterns across diverse forecasting horizons. Thirdly, a series-aware graph learning mechanism (SAGLM) which includes graph token encoding, adaptive adjacency matrix learning, and graph convolution, is proposed to integrate graph neural networks (GNNs) with Transformer within a synergistic framework to comprehensively capture intra-series and inter-series dependencies in net load datasets. Finally, case studies on the real-world datasets demonstrate that the PISAGT model can achieve a 14.4% improvement in prediction accuracy compared with direct forecasting methods, showing its effectiveness in net load forecasting.

A Data-Model Jointly Driven Framework for Visible Light Positioning Using Harmonic-Enhanced Graph Neural Networks

Visible light positioning (VLP), due to its widespread infrastructure deployment and high accuracy, has emerged as a highly promising key technology for Internet of Things (IoT). Current research mainly relies on either data-driven methods based on fingerprint features or model-driven methods based on geometric localization to estimate position. Although data-driven approaches can effectively cope with complex environmental disturbances, their performance is limited by insufficient exploitation of latent signal features on the one hand and strong dependence on training data on the other, resulting in limited generalization capability. In contrast, model-driven methods can adapt to different scenarios by leveraging physical models and geometric constraints, but they struggle to characterize complex interference patterns in dynamic environments. To address these issues, this article proposes a data-model jointly driven VLP framework that integrates the complementary strengths of both paradigms. First, at the framework level, a tightly coupled joint optimization scheme is constructed to integrate data-driven ranging with model-driven localization, preserving physical interpretability while leveraging the representation capability of deep learning. In contrast to conventional methods that discard harmonics as detrimental components, this article introduces a harmonic-enhanced ranging module that uses selected harmonic components as auxiliary structured spectral cues to improve the robustness of VLP ranging. The fundamental received signal strength (RSS) and selected harmonic RSS values of the modulation signal are incorporated into the ranging process, and a graph neural network (GNN)-based data-driven ranging model is developed to capture the structured relationships between the fundamental component and its harmonics, thereby improving ranging robustness in complex environments. Finally, to enable end-to-end optimization of the joint framework, a geometry-aware loss function is designed, allowing the model to jointly consider data fitting and physical geometric constraints during training, thereby coupling learned signal representations with model-driven geometric constraints.

A literature review of recurrent neural network approaches to malaria outbreak prediction in Sub-Saharan Africa

The review points out four persistent gaps: the lack of attention-augmented RNNs for malaria forecasting in Southern Africa, limited integration of health-facility infrastructure features with climate predictors, inadequate handling of missing data in African satellite-derived climate series, and the absence of operational dashboards that make model outputs usable for district health officers without statistical training.

Dynamic routing and wavelength assignment in optical networks: from capacity limits to flow-based control

Graph Neural Networks for Diffusion and Aggregation in Wireless Federated Learning

User devices (UDs) with non-independent and identically distributed (non-IID) data will worsen accuracy performance of the global model in federated learning (FL). Therefore, the implementation of diffusion strategies in machine learning (ML) models can enhance the effectiveness of federated learning with non-IID data. However, in a device-to-device (D2D) wireless federated learning (WFL) system, limited wireless resources and severe wireless channel interference become the important bottleneck to restrict the diffusion performance and model aggregation so as the global model of WFL with non-IID suffers from the weight divergence challenge. Thus, we propose a novel joint over-the-air computation (OAC) aggregation and diffusion framework by using a graph neural network (GNN) for WFL, termed an OAC-GNN-Dif framework. By integrating the OAC with message passing neural network (MPNN) of GNN, we further develop the OAC-MPNN-Dif algorithm based on the OAC-GNN-Dif framework. To further reduce communication costs, we designed an OAC message recurrent neural network (OAC-MPRNN-Dif) algorithm, where each UD propagates local models via D2D communications to refresh the graph embedding in the current frame based on the graph feature extraction and localization state of the previous frame to reduce communication costs. Additionally, we introduce dynamic time-varying MPNN for federated diffusion within evolving D2D network topologies. The experimental results indicate that our approach significantly performs well in communication overhead, with a 30%-60% decreasing in wireless resources overhead and 1.2-3.5 times decreasing in the number of model transfers compared to the FedDif methods. Moreover, our approach also improves the global model test accuracy, which is about 2.7% higher than the existing communication diffusion FL with non-IID characteristics.